Spaces:
Running
Running
File size: 50,199 Bytes
a045435 473cc8a a045435 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 | <!doctype html><html><head><meta charset='utf-8'><title>How humans push the T: measurements from the Push-T demonstrations</title><style>
body{max-width:980px;margin:2rem auto;padding:0 1.25rem;font:15px/1.55 -apple-system,Segoe UI,Roboto,sans-serif;color:#1c2330;background:#fafbfc}
h1{font-size:1.7rem;border-bottom:2px solid #d8dee8;padding-bottom:.4rem}h2{margin-top:2.2rem;border-bottom:1px solid #e3e8ef;padding-bottom:.25rem}
table{border-collapse:collapse;margin:1rem 0;font-size:13.5px;font-variant-numeric:tabular-nums}th,td{border:1px solid #d8dee8;padding:.35rem .6rem;text-align:left}
th{background:#eef2f7}td:not(:first-child){text-align:right}code{background:#eef2f7;padding:.1rem .3rem;border-radius:3px;font-size:.92em}
pre{background:#1c2330;color:#e6edf5;padding:.8rem 1rem;border-radius:6px;overflow:auto}pre code{background:none;color:inherit}
img{max-width:100%;border:1px solid #d8dee8;border-radius:4px;margin:.5rem 0}blockquote{border-left:3px solid #6aa9ff;margin:0;padding:.2rem 1rem;color:#4b5563}
nav{font-size:13px;margin-bottom:1.5rem}nav a{margin-right:1rem}
</style></head><body><nav><a href="index.html">index</a> <a href="brain3_vs_3a_events.html">brain3 vs 3a events</a> <a href="contact_friction_study.html">contact friction study</a> <a href="convergence_study.html">convergence study</a> <a href="dataset_definition.html">dataset definition</a> <a href="ds08_distribution_at_1000.html">ds08 distribution at 1000</a> <a href="ds_distribution_at_1000.html">ds distribution at 1000</a> <a href="ds_distribution_report.html">ds distribution report</a> <a href="ds_mu03_brain3b_distribution.html">ds mu03 brain3b distribution</a> <a href="episode_store.html">episode store</a> <a href="human_pushing_analysis.html">human pushing analysis</a> <a href="lewm_parity.html">lewm parity</a> <a href="reference_parity.html">reference parity</a> <a href="session_fidelity_and_behaviour.html">session fidelity and behaviour</a></nav><h1 id="how-humans-push-the-t-measurements-from-the-push-t-demonstrations">How humans push the T: measurements from the Push-T demonstrations</h1>
<p>Generated by <code>scripts/analyze_human_pushing.py</code> from <code>kp_file0.parquet</code>; every number and figure below is that script's output. Reproduce with:</p>
<pre><code>.venv/bin/python scripts/analyze_human_pushing.py /tmp/kp_file0.parquet
</code></pre>
<h2 id="data-and-provenance">Data and provenance</h2>
<ul>
<li>Source: LeRobot <code>pusht_keypoints</code>, file 0: 206 episodes, 25650 frames at 10 Hz. Columns used: <code>observation.state</code> (agent centre), <code>observation.environment_state</code> (8 block keypoints), <code>next.reward</code>, <code>next.success</code>.</li>
<li>These are the original human teleoperation demonstrations (mouse-driven pusher, Diffusion Policy's Push-T). The LeWM paper's dataset (<code>galilai-group/lewm-pusht</code>) is the DINO-WM Push-T set, which is these same demonstrations replayed with action noise (DINO-WM App. A.1, LeWM App. E). Velocity, acceleration and jerk must therefore be measured on the clean demonstrations, as done here; the noised replays would corrupt them.</li>
<li>Block pose per frame is recovered by a rigid fit of the 8 keypoints to the local vertices (bar <code>(-60,0) (60,0) (60,30) (-60,30)</code>, stem <code>(-15,30) (15,30) (15,120) (-15,120)</code>); max fit residual over the file: 3.8e-05 px.</li>
<li>Reward check: <code>next.reward[t]</code> equals <code>clip(coverage[t+1] / 0.95, 0, 1)</code> with coverage computed by <code>pusht_sim.envs.geometry.coverage</code> from the recovered pose; max deviation over the analysed episodes 2.2e-07. "Final coverage" below is <code>0.95 x next.reward</code> of the last frame, i.e. the coverage of the terminal state.</li>
<li><code>next.success</code> is never True in this file: no demonstration reaches 0.95 coverage before it ends (max terminal coverage 0.901).</li>
</ul>
<h2 id="definitions">Definitions</h2>
<ul>
<li><strong>Frames.</strong> World: 512 x 512 px, origin top-left. Block vertex frame: x along the bar, y from the bar's outer (top) edge (y = 0) through the bar (0..30) to the stem tip (y = 120); CoG at (0, 45); R = +x side, L = -x side. Goal pose (256, 256, 45 deg).</li>
<li><strong>Contact.</strong> <code>gap = dist(agent centre, T outline) - 15</code> (negative when the centre is inside the outline). Contact iff <code>gap <= 1.0 px</code>. The gap histogram over all 25650 frames has a spike in (-1, 0.5] px (Chipmunk's collision slop) and a flat tail beyond, so the threshold is not sensitive between 0.5 and 1.5 px.</li>
<li><strong>Contact point / edge / normal.</strong> Nearest point of the outline to the agent centre, in the vertex frame; the edge is the nearest of the 8 outline segments (<code>bar_top</code> y=0, <code>bar_end_R/L</code> x=+-60, <code>bar_bottom_R/L</code> y=30 beside the stem, <code>stem_R/L</code> x=+-15, <code>stem_tip</code> y=120); the outward normal is the unit vector from that point to the agent centre (at corners this is the corner-to-centre direction).</li>
<li><strong>Kinematics.</strong> Forward differences at DT = 0.1 s: <code>v[t] = (p[t+1]-p[t])/DT</code>, <code>a[t] = (v[t+1]-v[t])/DT</code>, <code>j[t] = (a[t+1]-a[t])/DT</code>; the report uses <code>|a|</code> and <code>|j|</code>. Block CoG velocity and rotation rate <code>omega</code> likewise from the recovered pose. Units px/s, px/s^2, px/s^3, rad/s (deg/s in tables).</li>
<li><strong>Roles (per frame).</strong> Let <code>v_n = -v . n_out</code> (speed into the block) and <code>v_t = |v . tangent|</code> at the nearest outline point; <code>dgap = (gap[t+1]-gap[t])/DT</code>.</li>
<li><code>TOUCH</code>: first frame of a contact bout (<code>contact[t] and not contact[t-1]</code>).</li>
<li><code>PUSH</code>: in contact, not TOUCH, <code>speed >= 10</code> px/s and <code>v_n >= v_t</code> (moving into the block, within 45 deg of the inward normal).</li>
<li><code>HOLD</code>: in contact, not TOUCH, <code>speed < 10</code> px/s (pausing while touching; kept separate so PUSH/SLIDE speeds are not diluted).</li>
<li><code>RETREAT</code> (release step): in contact, not TOUCH, <code>speed >= 10</code> px/s and <code>-v_n >= v_t</code> (moving away from the block faster than along it: this is the step that breaks contact).</li>
<li><code>SLIDE</code>: in contact, not TOUCH, <code>speed >= 10</code> px/s and <code>v_t > |v_n|</code> (moving along the outline).</li>
<li><code>APPROACH</code>: no contact, <code>speed >= 10</code> px/s and closing speed <code>-dgap >= sqrt(speed^2 - dgap^2)</code>, i.e. heading within 45 deg of the nearest outline point. The step that makes contact (<code>contact[t+1]</code>) is always APPROACH: contact resolution shortens its normal component, so its heading is not meaningful.</li>
<li><code>RETREAT</code>: no contact, <code>speed >= 10</code> px/s and <code>dgap >= sqrt(speed^2 - dgap^2)</code>, i.e. heading within 45 deg of straight away from the outline.</li>
<li><code>REPLAN/IDLE</code>: no contact and either <code>speed < 10</code> px/s (idle) or heading 45..135 deg off the outline direction (orbiting the block at roughly constant range to reach another edge). The same 45-deg normal-vs-tangential split is used in and out of contact.</li>
<li>The heading of no-contact motion (angle between the velocity and the direction to the nearest outline point) is unimodal, so any split is a convention; its distribution is reported in section 2 so the reader can re-cut it.</li>
<li>The last frame of an episode has no forward difference; it inherits the previous label (HOLD if that was TOUCH) or REPLAN/IDLE if not in contact.</li>
<li><strong>Bouts.</strong> Maximal runs of contact frames. A touch is <em>deliberate</em> if it starts the episode's first bout or follows >= 5 non-contact frames; otherwise it is a <em>re-contact</em> (chatter within one push).</li>
<li><strong>Errors.</strong> Position error = |goal CoG - block CoG| (px); angle error = wrap(goal angle - block angle) in (-180, 180] deg, positive = block must rotate in the +angle direction of the pose. Direction to goal is expressed in the block frame (<code>phi_g</code>, 0 deg = +x bar end, 90 = stem tip, -90 = bar top).</li>
<li><strong>Dominant error.</strong> Angle-dominant when <code>|angle error (rad)| x 60 px > position error</code>, i.e. the angle error moves the bar ends farther than the CoG is off; position-dominant otherwise.</li>
<li><strong>Lever arm.</strong> <code>l = (q - CoG) x (-n_out)</code> at the contact point q: the perpendicular offset of an inward-normal push from the CoG, px; sign = sign of the torque it produces. <code>|l| <= 15</code> px (one pusher radius) counts as a push through the CoG. A velocity-based variant uses the agent's velocity direction instead of the normal.</li>
<li><strong>Bearings.</strong> <code>bearing_block</code>: angle of the agent about the CoG in the block frame (same convention as <code>phi_g</code>). <code>bearing_goal</code>: signed angle between the goal->block axis and the CoG->agent vector; 0 = the agent is directly behind the block as seen from the goal (the pushing position), +-180 = between block and goal.</li>
</ul>
<h2 id="1-episode-selection">1. Episode selection</h2>
<p>206 of 206 episodes contain at least one contact frame; all analysis below uses the first 50 qualifying episodes by <code>episode_index</code> (0..49); the remaining 156 qualifying episodes are used only as an external hold-out for the contact-selection rule (section 6). No stratification by success is possible because <code>next.success</code> is never true.</p>
<table>
<thead>
<tr>
<th></th>
<th>n</th>
<th>mean</th>
<th>median</th>
<th>p10</th>
<th>p90</th>
</tr>
</thead>
<tbody>
<tr>
<td>episode length (frames)</td>
<td>50</td>
<td>125</td>
<td>126</td>
<td>83</td>
<td>159</td>
</tr>
<tr>
<td>final coverage</td>
<td>50</td>
<td>0.849</td>
<td>0.854</td>
<td>0.821</td>
<td>0.874</td>
</tr>
<tr>
<td>max coverage seen in episode</td>
<td>50</td>
<td>0.850</td>
<td>0.855</td>
<td>0.821</td>
<td>0.875</td>
</tr>
<tr>
<td>initial position error (px)</td>
<td>50</td>
<td>118.6</td>
<td>122.4</td>
<td>49.5</td>
<td>175.4</td>
</tr>
<tr>
<td>initial</td>
<td>angle error</td>
<td>(deg)</td>
<td>50</td>
<td>89.2</td>
<td>88.6</td>
</tr>
<tr>
<td>contact bouts per episode</td>
<td>50</td>
<td>3.1</td>
<td>3.0</td>
<td>2.0</td>
<td>5.0</td>
</tr>
<tr>
<td>contact frames per episode</td>
<td>50</td>
<td>62.4</td>
<td>60.0</td>
<td>40.0</td>
<td>90.3</td>
</tr>
</tbody>
</table>
<p><img alt="episodes" src="../figures/human_pushing/episode_overview.png" /></p>
<p>Per episode:</p>
<table>
<thead>
<tr>
<th>episode</th>
<th>frames</th>
<th>success</th>
<th>final coverage</th>
<th>bouts</th>
<th>role sequence (run-length compressed)</th>
</tr>
</thead>
<tbody>
<tr>
<td>0</td>
<td>161</td>
<td>no</td>
<td>0.830</td>
<td>3</td>
<td>R-I-A-T-P-S-P-S-I-R-I-A-T-P-H-P-H-P-H-S-I-A-T-P</td>
</tr>
<tr>
<td>1</td>
<td>118</td>
<td>no</td>
<td>0.855</td>
<td>2</td>
<td>I-A-T-P-S-P-S-P-S-P-H-S-I-R-I-A-T-H</td>
</tr>
<tr>
<td>2</td>
<td>141</td>
<td>no</td>
<td>0.860</td>
<td>2</td>
<td>R-I-A-T-P-S-P-S-P-S-P-S-P-S-H-S-I-A-T-P-S-R-I-A-I</td>
</tr>
<tr>
<td>3</td>
<td>159</td>
<td>no</td>
<td>0.874</td>
<td>3</td>
<td>R-I-A-T-P-S-I-R-I-A-T-P-S-R-I-A-T-P-S-P-S-P</td>
</tr>
<tr>
<td>4</td>
<td>159</td>
<td>no</td>
<td>0.864</td>
<td>4</td>
<td>I-R-I-A-T-P-H-P-H-R-I-A-T-P-R-I-A-I-A-T-P-H-S-I-A-T-P</td>
</tr>
<tr>
<td>5</td>
<td>157</td>
<td>no</td>
<td>0.853</td>
<td>2</td>
<td>I-A-T-P-S-P-R-I-A-I-A-T-S-P-H-R-I-A-I</td>
</tr>
<tr>
<td>6</td>
<td>69</td>
<td>no</td>
<td>0.845</td>
<td>2</td>
<td>I-A-T-P-S-R-I-A-T</td>
</tr>
<tr>
<td>7</td>
<td>169</td>
<td>no</td>
<td>0.854</td>
<td>6</td>
<td>R-I-A-T-P-S-I-A-T-I-A-T-P-S-H-S-I-A-T-P-S-R-I-A-T-P-S-R-I-A-T-P</td>
</tr>
<tr>
<td>8</td>
<td>80</td>
<td>no</td>
<td>0.831</td>
<td>3</td>
<td>I-A-I-A-T-P-H-R-I-A-T-P-S-P-S-I-R-I-A-T</td>
</tr>
<tr>
<td>9</td>
<td>134</td>
<td>no</td>
<td>0.866</td>
<td>3</td>
<td>I-A-I-A-I-A-T-P-S-P-S-I-A-T-P-H-S-R-I-A-I-A-T-P</td>
</tr>
<tr>
<td>10</td>
<td>188</td>
<td>no</td>
<td>0.873</td>
<td>4</td>
<td>I-A-I-A-T-P-S-P-H-S-I-A-T-P-H-S-P-S-I-R-I-A-T-P-H-R-I-A-T-P</td>
</tr>
<tr>
<td>11</td>
<td>149</td>
<td>no</td>
<td>0.863</td>
<td>2</td>
<td>A-I-R-I-A-I-A-I-A-T-P-S-P-S-P-S-P-S-P-S-P-S-P-H-S-I-A-T-P</td>
</tr>
<tr>
<td>12</td>
<td>140</td>
<td>no</td>
<td>0.855</td>
<td>5</td>
<td>I-R-I-A-T-P-S-I-A-T-P-S-I-A-T-P-S-R-I-A-T-S-P-H-R-I-A-T-H</td>
</tr>
<tr>
<td>13</td>
<td>125</td>
<td>no</td>
<td>0.864</td>
<td>3</td>
<td>I-R-I-A-T-P-R-I-A-T-P-S-P-S-R-I-A-T-P</td>
</tr>
<tr>
<td>14</td>
<td>107</td>
<td>no</td>
<td>0.834</td>
<td>3</td>
<td>A-I-R-I-A-I-A-T-P-S-P-H-S-I-A-T-P-S-I-A-T</td>
</tr>
<tr>
<td>15</td>
<td>153</td>
<td>no</td>
<td>0.797</td>
<td>6</td>
<td>A-I-A-T-P-S-I-A-T-P-S-R-I-A-T-P-R-I-A-T-P-S-R-I-A-T-P-H-P-S-I-R-I-A-I-A-T</td>
</tr>
<tr>
<td>16</td>
<td>126</td>
<td>no</td>
<td>0.841</td>
<td>3</td>
<td>I-A-T-P-S-I-R-I-A-T-P-H-R-I-A-T-P-S-P-S-P</td>
</tr>
<tr>
<td>17</td>
<td>102</td>
<td>no</td>
<td>0.827</td>
<td>2</td>
<td>I-A-T-S-P-S-P-S-P-R-I-A-T-P-S-P-S-H-P</td>
</tr>
<tr>
<td>18</td>
<td>86</td>
<td>no</td>
<td>0.860</td>
<td>2</td>
<td>I-A-I-A-T-P-S-P-S-I-A-T-P-H-P-S-P</td>
</tr>
<tr>
<td>19</td>
<td>75</td>
<td>no</td>
<td>0.836</td>
<td>1</td>
<td>A-I-A-T-P-S-P-S-P</td>
</tr>
<tr>
<td>20</td>
<td>142</td>
<td>no</td>
<td>0.867</td>
<td>3</td>
<td>I-R-I-A-T-P-S-P-S-P-H-P-S-I-R-I-A-I-A-T-P-R-I-A-T</td>
</tr>
<tr>
<td>21</td>
<td>115</td>
<td>no</td>
<td>0.875</td>
<td>2</td>
<td>I-A-T-P-H-P-R-I-A-T-P</td>
</tr>
<tr>
<td>22</td>
<td>109</td>
<td>no</td>
<td>0.875</td>
<td>4</td>
<td>R-I-A-T-P-S-I-A-T-P-S-I-A-T-P-S-R-I-A-T-P</td>
</tr>
<tr>
<td>23</td>
<td>110</td>
<td>no</td>
<td>0.843</td>
<td>3</td>
<td>I-A-T-P-S-R-I-A-T-P-H-S-R-I-A-T-P</td>
</tr>
<tr>
<td>24</td>
<td>98</td>
<td>no</td>
<td>0.835</td>
<td>3</td>
<td>I-A-I-A-T-P-S-R-I-A-T-P-H-P-S-I-A-T-P</td>
</tr>
<tr>
<td>25</td>
<td>133</td>
<td>no</td>
<td>0.870</td>
<td>2</td>
<td>A-R-I-A-T-P-S-P-S-P-S-I-A-T-P</td>
</tr>
<tr>
<td>26</td>
<td>158</td>
<td>no</td>
<td>0.845</td>
<td>5</td>
<td>I-A-I-A-T-P-S-P-S-R-I-A-T-S-P-S-A-T-P-S-R-I-A-T-P-H-R-I-A-T</td>
</tr>
<tr>
<td>27</td>
<td>112</td>
<td>no</td>
<td>0.820</td>
<td>3</td>
<td>I-A-I-A-I-A-T-P-H-P-S-I-A-I-A-T-P-H-P-S-I-R-I-A-T-P</td>
</tr>
<tr>
<td>28</td>
<td>107</td>
<td>no</td>
<td>0.847</td>
<td>2</td>
<td>I-A-T-P-H-P-R-I-A-T-P</td>
</tr>
<tr>
<td>29</td>
<td>115</td>
<td>no</td>
<td>0.821</td>
<td>2</td>
<td>I-A-I-A-T-S-P-S-P-S-P-S-R-I-A-T-P</td>
</tr>
<tr>
<td>30</td>
<td>84</td>
<td>no</td>
<td>0.842</td>
<td>2</td>
<td>I-A-T-P-S-P-R-I-A-T-P</td>
</tr>
<tr>
<td>31</td>
<td>95</td>
<td>no</td>
<td>0.892</td>
<td>3</td>
<td>I-R-I-A-T-P-S-I-A-T-S-P-S-P-S-P-S-P-H-S-I-A-T-P</td>
</tr>
<tr>
<td>32</td>
<td>148</td>
<td>no</td>
<td>0.901</td>
<td>4</td>
<td>A-I-A-I-A-T-P-S-P-S-R-I-A-T-P-H-S-I-A-T-P-S-I-A-T-H</td>
</tr>
<tr>
<td>33</td>
<td>120</td>
<td>no</td>
<td>0.866</td>
<td>3</td>
<td>A-R-A-I-A-T-P-H-R-I-A-T-P-S-I-A-T-P</td>
</tr>
<tr>
<td>34</td>
<td>142</td>
<td>no</td>
<td>0.855</td>
<td>4</td>
<td>I-A-I-A-T-P-S-P-S-P-S-R-I-A-T-P-S-I-R-I-A-I-A-T-P-S-R-I-A-T</td>
</tr>
<tr>
<td>35</td>
<td>143</td>
<td>no</td>
<td>0.852</td>
<td>2</td>
<td>R-I-A-I-A-T-S-P-S-P-S-P-S-R-I-A-T-P</td>
</tr>
<tr>
<td>36</td>
<td>142</td>
<td>no</td>
<td>0.861</td>
<td>3</td>
<td>A-I-A-T-P-S-P-H-S-I-R-I-A-I-A-I-A-T-P-R-I-A-T-P-H</td>
</tr>
<tr>
<td>37</td>
<td>121</td>
<td>no</td>
<td>0.831</td>
<td>3</td>
<td>I-A-I-A-T-P-S-P-S-I-A-T-P-S-R-I-A-T-P</td>
</tr>
<tr>
<td>38</td>
<td>186</td>
<td>no</td>
<td>0.859</td>
<td>7</td>
<td>I-R-I-A-I-A-I-A-T-P-S-I-A-T-P-S-I-A-T-P-S-P-S-R-I-A-T-P-H-P-H-R-I-A-T-P-S-I-A-T-P-R-I-A-T-P</td>
</tr>
<tr>
<td>39</td>
<td>99</td>
<td>no</td>
<td>0.838</td>
<td>2</td>
<td>R-I-A-I-A-T-P-S-P-S-P-S-P-S-P-S-P-S-R-I-A-T-P</td>
</tr>
<tr>
<td>40</td>
<td>139</td>
<td>no</td>
<td>0.870</td>
<td>3</td>
<td>I-A-I-A-T-P-S-P-S-P-S-P-S-P-S-R-I-A-T-P-S-P-H-R-I-A-T-S-P</td>
</tr>
<tr>
<td>41</td>
<td>73</td>
<td>no</td>
<td>0.835</td>
<td>2</td>
<td>R-A-T-P-H-P-H-P-R-I-A-T-P</td>
</tr>
<tr>
<td>42</td>
<td>136</td>
<td>no</td>
<td>0.782</td>
<td>4</td>
<td>I-A-I-A-T-P-H-R-I-A-T-P-S-P-S-R-I-A-T-P-S-I-A-T</td>
</tr>
<tr>
<td>43</td>
<td>118</td>
<td>no</td>
<td>0.888</td>
<td>6</td>
<td>T-P-S-I-R-I-A-T-S-I-A-T-P-R-I-A-T-I-A-T-P-S-P-S-P-S-P-S-P-S-I-A-I-A-T-P</td>
</tr>
<tr>
<td>44</td>
<td>127</td>
<td>no</td>
<td>0.866</td>
<td>3</td>
<td>I-A-T-P-S-R-I-A-I-A-T-P-S-H-S-P-H-S-I-A-T-P</td>
</tr>
<tr>
<td>45</td>
<td>152</td>
<td>no</td>
<td>0.790</td>
<td>4</td>
<td>I-A-T-P-S-I-R-I-A-T-P-H-S-I-A-I-A-T-P-H-P-R-I-A-T-H</td>
</tr>
<tr>
<td>46</td>
<td>83</td>
<td>no</td>
<td>0.862</td>
<td>2</td>
<td>I-R-I-A-T-P-S-P-S-P-S-P-S-P-S-P-S-P-S-I-A-T-P</td>
</tr>
<tr>
<td>47</td>
<td>109</td>
<td>no</td>
<td>0.855</td>
<td>2</td>
<td>I-A-I-A-I-A-T-P-S-P-R-I-A-T-P-S-P-S-P</td>
</tr>
<tr>
<td>48</td>
<td>138</td>
<td>no</td>
<td>0.838</td>
<td>3</td>
<td>I-R-I-A-I-A-T-S-P-S-I-A-T-S-P-S-P-R-S-I-A-T-P</td>
</tr>
<tr>
<td>49</td>
<td>83</td>
<td>no</td>
<td>0.796</td>
<td>2</td>
<td>I-A-I-A-T-P-S-R-I-A-T-P</td>
</tr>
</tbody>
</table>
<h2 id="2-phase-segmentation">2. Phase segmentation</h2>
<p>Per-frame labels are in <code>figures/human_pushing/frame_labels.csv</code>. Frame shares and run lengths (a run = maximal sequence of one role):</p>
<table>
<thead>
<tr>
<th>role</th>
<th>frames</th>
<th>share</th>
<th>runs</th>
<th>mean run (frames)</th>
<th>median</th>
<th>p10</th>
<th>p90</th>
</tr>
</thead>
<tbody>
<tr>
<td>APPROACH</td>
<td>829</td>
<td>13.3%</td>
<td>202</td>
<td>4.1</td>
<td>4</td>
<td>1</td>
<td>8</td>
</tr>
<tr>
<td>TOUCH</td>
<td>154</td>
<td>2.5%</td>
<td>154</td>
<td>1.0</td>
<td>1</td>
<td>1</td>
<td>1</td>
</tr>
<tr>
<td>PUSH</td>
<td>2404</td>
<td>38.6%</td>
<td>232</td>
<td>10.4</td>
<td>7</td>
<td>2</td>
<td>25</td>
</tr>
<tr>
<td>SLIDE</td>
<td>416</td>
<td>6.7%</td>
<td>169</td>
<td>2.5</td>
<td>2</td>
<td>1</td>
<td>5</td>
</tr>
<tr>
<td>RETREAT</td>
<td>170</td>
<td>2.7%</td>
<td>88</td>
<td>1.9</td>
<td>2</td>
<td>1</td>
<td>3</td>
</tr>
<tr>
<td>REPLAN/IDLE</td>
<td>2156</td>
<td>34.6%</td>
<td>216</td>
<td>10.0</td>
<td>8</td>
<td>2</td>
<td>19</td>
</tr>
<tr>
<td>HOLD</td>
<td>106</td>
<td>1.7%</td>
<td>49</td>
<td>2.2</td>
<td>1</td>
<td>1</td>
<td>4</td>
</tr>
</tbody>
</table>
<ul>
<li>REPLAN/IDLE splits into 19 idle frames (speed < 10 px/s) and 2137 orbiting frames (moving around the block at 45..135 deg to the outline direction).</li>
<li>Heading of no-contact motion relative to the nearest outline point (n = 3093 frames at >= 10 px/s): median 74 deg, p10 18, p90 119; 24% within 45 deg of the block (APPROACH), 72% tangential (orbit), 4% within 45 deg of straight away (RETREAT).</li>
<li>The block is moving (CoG >= 5 px/s or |omega| >= 2 deg/s) in 98.5% of PUSH frames and 74.8% of SLIDE frames; it is rotating in 64.7% of SLIDE frames.</li>
<li>Contact bouts per episode: mean 3.08, median 3, p10 2, p90 5. Bout length (frames): median 17, p10 3, p90 40. Time between bouts (non-contact frames): median 16, p10 7, p90 25 (n = 104); 5 of these gaps are shorter than 5 frames (re-contacts).</li>
<li>Frames before the first touch: median 27, p10 14, p90 42.</li>
</ul>
<p>Transitions between consecutive runs (row = from, column = to, row-normalised %):</p>
<table>
<thead>
<tr>
<th>from \ to</th>
<th>APPROACH</th>
<th>TOUCH</th>
<th>PUSH</th>
<th>SLIDE</th>
<th>RETREAT</th>
<th>REPLAN/IDLE</th>
<th>HOLD</th>
</tr>
</thead>
<tbody>
<tr>
<td>APPROACH (n=202)</td>
<td>0</td>
<td>76</td>
<td>0</td>
<td>0</td>
<td>1</td>
<td>23</td>
<td>0</td>
</tr>
<tr>
<td>TOUCH (n=146)</td>
<td>0</td>
<td>0</td>
<td>88</td>
<td>8</td>
<td>0</td>
<td>1</td>
<td>3</td>
</tr>
<tr>
<td>PUSH (n=197)</td>
<td>0</td>
<td>0</td>
<td>0</td>
<td>71</td>
<td>8</td>
<td>0</td>
<td>21</td>
</tr>
<tr>
<td>SLIDE (n=169)</td>
<td>1</td>
<td>0</td>
<td>51</td>
<td>0</td>
<td>17</td>
<td>29</td>
<td>2</td>
</tr>
<tr>
<td>RETREAT (n=88)</td>
<td>2</td>
<td>0</td>
<td>0</td>
<td>1</td>
<td>0</td>
<td>97</td>
<td>0</td>
</tr>
<tr>
<td>REPLAN/IDLE (n=214)</td>
<td>89</td>
<td>0</td>
<td>0</td>
<td>0</td>
<td>11</td>
<td>0</td>
<td>0</td>
</tr>
<tr>
<td>HOLD (n=44)</td>
<td>0</td>
<td>0</td>
<td>36</td>
<td>39</td>
<td>25</td>
<td>0</td>
<td>0</td>
</tr>
</tbody>
</table>
<p>Most frequent run trigrams (A=APPROACH, T=TOUCH, P=PUSH, S=SLIDE, R=RETREAT, I=REPLAN/IDLE, H=HOLD):</p>
<table>
<thead>
<tr>
<th>trigram</th>
<th>count</th>
</tr>
</thead>
<tbody>
<tr>
<td>I-A-T</td>
<td>151</td>
</tr>
<tr>
<td>A-T-P</td>
<td>128</td>
</tr>
<tr>
<td>R-I-A</td>
<td>85</td>
</tr>
<tr>
<td>P-S-P</td>
<td>75</td>
</tr>
<tr>
<td>T-P-S</td>
<td>69</td>
</tr>
<tr>
<td>S-P-S</td>
<td>65</td>
</tr>
<tr>
<td>A-I-A</td>
<td>43</td>
</tr>
<tr>
<td>I-A-I</td>
<td>40</td>
</tr>
<tr>
<td>S-I-A</td>
<td>36</td>
</tr>
<tr>
<td>P-S-I</td>
<td>34</td>
</tr>
<tr>
<td>S-R-I</td>
<td>28</td>
</tr>
<tr>
<td>P-S-R</td>
<td>26</td>
</tr>
</tbody>
</table>
<p>Within a bout, the first frame after TOUCH is PUSH in 90% of bouts longer than one frame, SLIDE in 8%. A bout contains PUSH in 90% of cases and SLIDE in 60%. 48 of the 154 bouts run to the end of the episode; for the 106 bouts that end in a release, the release step is SLIDE 61%, RETREAT 37%, TOUCH 2% (heading of the release step median 124 deg off the inward normal), and the first frame after release is REPLAN/IDLE 58%, RETREAT 41%, APPROACH 1%. The block is already at rest when contact is broken: block speed on the release step median 0.4 px/s, p90 2.2, below 5 px/s in 97% of releases; one frame later it is 0.0 px/s (median), so the block does not coast after release.</p>
<p><img alt="timeline" src="../figures/human_pushing/role_timeline.png" /></p>
<p><img alt="durations" src="../figures/human_pushing/role_durations.png" /></p>
<h2 id="3-kinematics-per-role">3. Kinematics per role</h2>
<p><strong>distance agent-outline minus radius (px)</strong></p>
<table>
<thead>
<tr>
<th></th>
<th>n</th>
<th>mean</th>
<th>median</th>
<th>p10</th>
<th>p90</th>
</tr>
</thead>
<tbody>
<tr>
<td>APPROACH</td>
<td>829</td>
<td>38.1</td>
<td>20.0</td>
<td>3.7</td>
<td>110.7</td>
</tr>
<tr>
<td>TOUCH</td>
<td>154</td>
<td>-0.2</td>
<td>-0.1</td>
<td>-0.5</td>
<td>0.6</td>
</tr>
<tr>
<td>PUSH</td>
<td>2404</td>
<td>-0.2</td>
<td>-0.1</td>
<td>-0.3</td>
<td>-0.0</td>
</tr>
<tr>
<td>SLIDE</td>
<td>416</td>
<td>-0.1</td>
<td>-0.1</td>
<td>-0.2</td>
<td>0.0</td>
</tr>
<tr>
<td>RETREAT</td>
<td>170</td>
<td>31.2</td>
<td>10.4</td>
<td>-0.0</td>
<td>87.1</td>
</tr>
<tr>
<td>REPLAN/IDLE</td>
<td>2156</td>
<td>46.8</td>
<td>36.8</td>
<td>10.5</td>
<td>92.8</td>
</tr>
<tr>
<td>HOLD</td>
<td>106</td>
<td>-0.0</td>
<td>-0.0</td>
<td>-0.1</td>
<td>-0.0</td>
</tr>
</tbody>
</table>
<p><img alt="gap" src="../figures/human_pushing/kin_gap.png" /></p>
<p><strong>heading relative to the block (deg; 0 = straight into the nearest outline point / inward normal, 90 = along it, 180 = straight away)</strong></p>
<table>
<thead>
<tr>
<th></th>
<th>n</th>
<th>mean</th>
<th>median</th>
<th>p10</th>
<th>p90</th>
</tr>
</thead>
<tbody>
<tr>
<td>APPROACH</td>
<td>829</td>
<td>26.2</td>
<td>24.5</td>
<td>4.1</td>
<td>44.7</td>
</tr>
<tr>
<td>TOUCH</td>
<td>144</td>
<td>20.8</td>
<td>15.6</td>
<td>3.3</td>
<td>51.3</td>
</tr>
<tr>
<td>PUSH</td>
<td>2369</td>
<td>18.3</td>
<td>16.1</td>
<td>2.9</td>
<td>36.7</td>
</tr>
<tr>
<td>SLIDE</td>
<td>416</td>
<td>70.7</td>
<td>63.6</td>
<td>47.3</td>
<td>103.2</td>
</tr>
<tr>
<td>RETREAT</td>
<td>170</td>
<td>150.7</td>
<td>147.7</td>
<td>137.1</td>
<td>168.7</td>
</tr>
<tr>
<td>REPLAN/IDLE</td>
<td>2135</td>
<td>84.7</td>
<td>82.9</td>
<td>55.8</td>
<td>116.7</td>
</tr>
<tr>
<td>HOLD</td>
<td>0</td>
<td>-</td>
<td>-</td>
<td>-</td>
<td>-</td>
</tr>
</tbody>
</table>
<p><img alt="heading" src="../figures/human_pushing/kin_heading.png" /></p>
<p><strong>bearing of agent about the CoG, block frame (deg; 0=+x bar end, 90=stem tip, -90=bar top)</strong></p>
<table>
<thead>
<tr>
<th></th>
<th>n</th>
<th>mean</th>
<th>median</th>
<th>p10</th>
<th>p90</th>
</tr>
</thead>
<tbody>
<tr>
<td>APPROACH</td>
<td>829</td>
<td>9.6</td>
<td>29.5</td>
<td>-155.3</td>
<td>146.7</td>
</tr>
<tr>
<td>TOUCH</td>
<td>154</td>
<td>3.8</td>
<td>17.0</td>
<td>-153.9</td>
<td>113.9</td>
</tr>
<tr>
<td>PUSH</td>
<td>2404</td>
<td>-21.8</td>
<td>-16.3</td>
<td>-179.4</td>
<td>115.1</td>
</tr>
<tr>
<td>SLIDE</td>
<td>416</td>
<td>-25.8</td>
<td>-17.0</td>
<td>-179.7</td>
<td>114.4</td>
</tr>
<tr>
<td>RETREAT</td>
<td>170</td>
<td>19.9</td>
<td>43.3</td>
<td>-140.3</td>
<td>150.5</td>
</tr>
<tr>
<td>REPLAN/IDLE</td>
<td>2156</td>
<td>-1.3</td>
<td>-2.7</td>
<td>-151.8</td>
<td>150.8</td>
</tr>
<tr>
<td>HOLD</td>
<td>106</td>
<td>-20.6</td>
<td>-0.1</td>
<td>-180.0</td>
<td>119.2</td>
</tr>
</tbody>
</table>
<p><img alt="bearing_block" src="../figures/human_pushing/kin_bearing_block.png" /></p>
<p><strong>bearing of agent off the goal->block axis (deg; 0=behind block, opposite the goal)</strong></p>
<table>
<thead>
<tr>
<th></th>
<th>n</th>
<th>mean</th>
<th>median</th>
<th>p10</th>
<th>p90</th>
</tr>
</thead>
<tbody>
<tr>
<td>APPROACH</td>
<td>829</td>
<td>10.0</td>
<td>3.6</td>
<td>-59.9</td>
<td>114.2</td>
</tr>
<tr>
<td>TOUCH</td>
<td>154</td>
<td>0.2</td>
<td>-0.1</td>
<td>-55.6</td>
<td>47.1</td>
</tr>
<tr>
<td>PUSH</td>
<td>2404</td>
<td>7.3</td>
<td>4.2</td>
<td>-72.0</td>
<td>89.3</td>
</tr>
<tr>
<td>SLIDE</td>
<td>416</td>
<td>5.3</td>
<td>8.0</td>
<td>-84.8</td>
<td>94.0</td>
</tr>
<tr>
<td>RETREAT</td>
<td>170</td>
<td>30.9</td>
<td>52.6</td>
<td>-117.6</td>
<td>158.5</td>
</tr>
<tr>
<td>REPLAN/IDLE</td>
<td>2156</td>
<td>7.9</td>
<td>10.8</td>
<td>-114.6</td>
<td>122.9</td>
</tr>
<tr>
<td>HOLD</td>
<td>106</td>
<td>44.1</td>
<td>68.7</td>
<td>-91.8</td>
<td>155.4</td>
</tr>
</tbody>
</table>
<p><img alt="bearing_goal" src="../figures/human_pushing/kin_bearing_goal.png" /></p>
<p><strong>agent speed (px/s)</strong></p>
<table>
<thead>
<tr>
<th></th>
<th>n</th>
<th>mean</th>
<th>median</th>
<th>p10</th>
<th>p90</th>
</tr>
</thead>
<tbody>
<tr>
<td>APPROACH</td>
<td>829</td>
<td>83.7</td>
<td>69.3</td>
<td>32.1</td>
<td>151.0</td>
</tr>
<tr>
<td>TOUCH</td>
<td>146</td>
<td>59.5</td>
<td>57.3</td>
<td>25.6</td>
<td>90.4</td>
</tr>
<tr>
<td>PUSH</td>
<td>2369</td>
<td>52.9</td>
<td>47.9</td>
<td>19.9</td>
<td>89.4</td>
</tr>
<tr>
<td>SLIDE</td>
<td>416</td>
<td>58.2</td>
<td>49.6</td>
<td>18.4</td>
<td>106.0</td>
</tr>
<tr>
<td>RETREAT</td>
<td>170</td>
<td>126.8</td>
<td>105.7</td>
<td>32.2</td>
<td>250.4</td>
</tr>
<tr>
<td>REPLAN/IDLE</td>
<td>2154</td>
<td>128.7</td>
<td>121.6</td>
<td>54.6</td>
<td>208.5</td>
</tr>
<tr>
<td>HOLD</td>
<td>101</td>
<td>6.4</td>
<td>6.5</td>
<td>2.9</td>
<td>9.1</td>
</tr>
</tbody>
</table>
<p><img alt="speed" src="../figures/human_pushing/kin_speed.png" /></p>
<p><strong>agent |acceleration| (px/s^2)</strong></p>
<table>
<thead>
<tr>
<th></th>
<th>n</th>
<th>mean</th>
<th>median</th>
<th>p10</th>
<th>p90</th>
</tr>
</thead>
<tbody>
<tr>
<td>APPROACH</td>
<td>819</td>
<td>194</td>
<td>147</td>
<td>50</td>
<td>362</td>
</tr>
<tr>
<td>TOUCH</td>
<td>142</td>
<td>120</td>
<td>104</td>
<td>40</td>
<td>207</td>
</tr>
<tr>
<td>PUSH</td>
<td>2334</td>
<td>121</td>
<td>88</td>
<td>29</td>
<td>230</td>
</tr>
<tr>
<td>SLIDE</td>
<td>416</td>
<td>280</td>
<td>145</td>
<td>43</td>
<td>730</td>
</tr>
<tr>
<td>RETREAT</td>
<td>170</td>
<td>706</td>
<td>567</td>
<td>224</td>
<td>1244</td>
</tr>
<tr>
<td>REPLAN/IDLE</td>
<td>2154</td>
<td>360</td>
<td>305</td>
<td>105</td>
<td>682</td>
</tr>
<tr>
<td>HOLD</td>
<td>100</td>
<td>216</td>
<td>47</td>
<td>15</td>
<td>764</td>
</tr>
</tbody>
</table>
<p><img alt="accel" src="../figures/human_pushing/kin_accel.png" /></p>
<p><strong>agent |jerk| (px/s^3)</strong></p>
<table>
<thead>
<tr>
<th></th>
<th>n</th>
<th>mean</th>
<th>median</th>
<th>p10</th>
<th>p90</th>
</tr>
</thead>
<tbody>
<tr>
<td>APPROACH</td>
<td>806</td>
<td>1592</td>
<td>1097</td>
<td>331</td>
<td>3105</td>
</tr>
<tr>
<td>TOUCH</td>
<td>141</td>
<td>1055</td>
<td>828</td>
<td>268</td>
<td>1896</td>
</tr>
<tr>
<td>PUSH</td>
<td>2302</td>
<td>1311</td>
<td>798</td>
<td>252</td>
<td>2625</td>
</tr>
<tr>
<td>SLIDE</td>
<td>415</td>
<td>2801</td>
<td>1316</td>
<td>313</td>
<td>7583</td>
</tr>
<tr>
<td>RETREAT</td>
<td>170</td>
<td>6032</td>
<td>4054</td>
<td>1727</td>
<td>12127</td>
</tr>
<tr>
<td>REPLAN/IDLE</td>
<td>2153</td>
<td>2444</td>
<td>1778</td>
<td>533</td>
<td>5131</td>
</tr>
<tr>
<td>HOLD</td>
<td>98</td>
<td>2481</td>
<td>708</td>
<td>144</td>
<td>6606</td>
</tr>
</tbody>
</table>
<p><img alt="jerk" src="../figures/human_pushing/kin_jerk.png" /></p>
<p>Block motion per role (block CoG speed px/s, |rotation rate| deg/s):</p>
<table>
<thead>
<tr>
<th></th>
<th>n</th>
<th>mean</th>
<th>median</th>
<th>p10</th>
<th>p90</th>
</tr>
</thead>
<tbody>
<tr>
<td>APPROACH block speed</td>
<td>829</td>
<td>2.0</td>
<td>0.0</td>
<td>0.0</td>
<td>4.8</td>
</tr>
<tr>
<td>TOUCH block speed</td>
<td>146</td>
<td>37.6</td>
<td>34.4</td>
<td>12.9</td>
<td>67.1</td>
</tr>
<tr>
<td>PUSH block speed</td>
<td>2369</td>
<td>35.8</td>
<td>32.4</td>
<td>13.7</td>
<td>60.6</td>
</tr>
<tr>
<td>SLIDE block speed</td>
<td>416</td>
<td>17.2</td>
<td>12.0</td>
<td>0.1</td>
<td>43.9</td>
</tr>
<tr>
<td>RETREAT block speed</td>
<td>170</td>
<td>0.2</td>
<td>0.0</td>
<td>0.0</td>
<td>0.6</td>
</tr>
<tr>
<td>REPLAN/IDLE block speed</td>
<td>2154</td>
<td>0.0</td>
<td>0.0</td>
<td>0.0</td>
<td>0.0</td>
</tr>
<tr>
<td>HOLD block speed</td>
<td>101</td>
<td>4.2</td>
<td>4.0</td>
<td>1.2</td>
<td>7.6</td>
</tr>
<tr>
<td>APPROACH</td>
<td>omega</td>
<td></td>
<td>829</td>
<td>1.0</td>
<td>0.0</td>
</tr>
<tr>
<td>TOUCH</td>
<td>omega</td>
<td></td>
<td>146</td>
<td>17.7</td>
<td>11.9</td>
</tr>
<tr>
<td>PUSH</td>
<td>omega</td>
<td></td>
<td>2369</td>
<td>18.1</td>
<td>15.4</td>
</tr>
<tr>
<td>SLIDE</td>
<td>omega</td>
<td></td>
<td>416</td>
<td>7.6</td>
<td>4.5</td>
</tr>
<tr>
<td>RETREAT</td>
<td>omega</td>
<td></td>
<td>170</td>
<td>0.1</td>
<td>0.0</td>
</tr>
<tr>
<td>REPLAN/IDLE</td>
<td>omega</td>
<td></td>
<td>2154</td>
<td>0.0</td>
<td>0.0</td>
</tr>
<tr>
<td>HOLD</td>
<td>omega</td>
<td></td>
<td>101</td>
<td>1.8</td>
<td>1.7</td>
</tr>
</tbody>
</table>
<h2 id="4-approach-geometry-at-touch">4. Approach geometry at TOUCH</h2>
<p>154 touches in the 50 episodes, 149 deliberate (first bout or >= 5 non-contact frames before) and 5 re-contacts. Statistics below are over deliberate touches; the approach velocity at touch is the last free-flight step before the contact step, <code>(p[t-1]-p[t-2])/DT</code> (the contact step itself is shortened along the normal by contact resolution and would bias the incidence towards glancing), and the incidence angle is measured between it and the inward normal at the contact point (0 = head-on, 90 = glancing).</p>
<table>
<thead>
<tr>
<th>edge</th>
<th>touches</th>
<th>share</th>
</tr>
</thead>
<tbody>
<tr>
<td>bar_top</td>
<td>39</td>
<td>26.2%</td>
</tr>
<tr>
<td>bar_end_R</td>
<td>7</td>
<td>4.7%</td>
</tr>
<tr>
<td>bar_bottom_R</td>
<td>4</td>
<td>2.7%</td>
</tr>
<tr>
<td>stem_R</td>
<td>25</td>
<td>16.8%</td>
</tr>
<tr>
<td>stem_tip</td>
<td>33</td>
<td>22.1%</td>
</tr>
<tr>
<td>stem_L</td>
<td>20</td>
<td>13.4%</td>
</tr>
<tr>
<td>bar_bottom_L</td>
<td>9</td>
<td>6.0%</td>
</tr>
<tr>
<td>bar_end_L</td>
<td>12</td>
<td>8.1%</td>
</tr>
<tr>
<td><em>stem_side</em> (L+R)</td>
<td>45</td>
<td>30.2%</td>
</tr>
<tr>
<td><em>bar_top</em> (L+R)</td>
<td>39</td>
<td>26.2%</td>
</tr>
<tr>
<td><em>stem_tip</em> (L+R)</td>
<td>33</td>
<td>22.1%</td>
</tr>
<tr>
<td><em>bar_end</em> (L+R)</td>
<td>19</td>
<td>12.8%</td>
</tr>
<tr>
<td><em>bar_bottom</em> (L+R)</td>
<td>13</td>
<td>8.7%</td>
</tr>
</tbody>
</table>
<table>
<thead>
<tr>
<th></th>
<th>n</th>
<th>mean</th>
<th>median</th>
<th>p10</th>
<th>p90</th>
</tr>
</thead>
<tbody>
<tr>
<td>incidence angle (deg)</td>
<td>149</td>
<td>23.6</td>
<td>17.1</td>
<td>2.0</td>
<td>51.3</td>
</tr>
<tr>
<td>approach speed at touch (px/s)</td>
<td>149</td>
<td>68.0</td>
<td>67.0</td>
<td>29.2</td>
<td>105.9</td>
</tr>
<tr>
<td>APPROACH run length before touch (frames)</td>
<td>149</td>
<td>4</td>
<td>4</td>
<td>1</td>
<td>8</td>
</tr>
<tr>
<td>standoff before re-approach: max gap between bouts (px)</td>
<td>99</td>
<td>42.0</td>
<td>38.8</td>
<td>16.7</td>
<td>70.1</td>
</tr>
<tr>
<td>bearing change around block, last contact -> next touch (deg)</td>
<td>99</td>
<td>-25.5</td>
<td>-28.4</td>
<td>-130.6</td>
<td>104.3</td>
</tr>
<tr>
<td></td>
<td>bearing change</td>
<td>(deg)</td>
<td>99</td>
<td>81.8</td>
<td>88.6</td>
</tr>
<tr>
<td>non-contact frames before touch (bouts after the first)</td>
<td>99</td>
<td>17</td>
<td>16</td>
<td>9</td>
<td>25</td>
</tr>
<tr>
<td></td>
<td>agent bearing off goal->block axis</td>
<td>at touch (deg, all bouts)</td>
<td>154</td>
<td>31.9</td>
<td>24.4</td>
</tr>
<tr>
<td></td>
<td>agent bearing off goal->block axis</td>
<td>at release (deg, released bouts)</td>
<td>106</td>
<td>85.4</td>
<td>85.5</td>
</tr>
</tbody>
</table>
<p>Head-on share (incidence <= 30 deg): 72%; glancing share (incidence >= 60 deg): 7%. The edge touched is the one nearest the agent at the previous release (or at the episode start) in 9% of deliberate touches (11% by edge type): the pusher usually travels to a different edge rather than touching the nearest one. Perimeter distance travelled around the outline from the previous release point to the next touch point: median 98 px, p10 25, p90 182 (perimeter 480 px). Gap at the start of the final APPROACH run: median 23 px, p10 3, p90 54.</p>
<p><img alt="touch" src="../figures/human_pushing/touch_geometry.png" /></p>
<p><img alt="overlay" src="../figures/human_pushing/contact_points_overlay.png" /></p>
<h2 id="5-push-dynamics">5. Push dynamics</h2>
<p>Over PUSH frames with a defined rotation rate (n = 2369): Pearson correlation between the inward-normal lever arm and the block rotation rate r = 0.869 (velocity-direction lever arm: r = 0.854); least-squares slope 0.562 deg/s per px of lever arm; the sign of the lever arm matches the sign of the rotation in 100% of PUSH frames where the block rotates faster than 2 deg/s.</p>
<p>| PUSH frames by dominant error | n | median |lever| (px) | through CoG (|l| <= 15 px) | corrective torque sign | median block speed (px/s) | median |omega| (deg/s) |
|---|---|---|---|---|---|---|
| position-dominant | 1883 | 25.2 | 40.4 | 78.5 | 31.4 | 11.5 |
| angle-dominant | 486 | 42.5 | 11.5 | 99.4 | 35.5 | 26.1 |
| all PUSH | 2369 | 30.1 | 34.4 | 82.8 | 32.4 | 15.4 |</p>
<p>| |angle error| bin (deg) | PUSH frames | median |lever| (px) | lever sign corrective | rotation sign corrective | median |omega| (deg/s) |
|---|---|---|---|---|---|
| 0-5 | 652 | 5.7 | 56.7 | 57.5 | 3.9 |
| 5-10 | 286 | 11.8 | 79.4 | 80.4 | 6.4 |
| 10-20 | 275 | 33.9 | 92.0 | 93.8 | 15.2 |
| 20-30 | 191 | 37.8 | 92.7 | 92.7 | 19.4 |
| 30-45 | 197 | 42.5 | 99.0 | 99.5 | 29.0 |
| 45-60 | 197 | 42.7 | 92.9 | 93.4 | 25.4 |
| 60-90 | 266 | 43.3 | 96.2 | 96.2 | 28.6 |
| 90-180 | 305 | 36.5 | 98.7 | 98.4 | 27.5 |</p>
<p>Per bout (all bouts, n = 154):</p>
<table>
<thead>
<tr>
<th></th>
<th>n</th>
<th>mean</th>
<th>median</th>
<th>p10</th>
<th>p90</th>
</tr>
</thead>
<tbody>
<tr>
<td>block CoG displacement per bout (px)</td>
<td>154</td>
<td>57.4</td>
<td>44.0</td>
<td>4.4</td>
<td>125.1</td>
</tr>
<tr>
<td>displacement toward goal per bout (px)</td>
<td>154</td>
<td>49.0</td>
<td>35.0</td>
<td>0.0</td>
<td>115.5</td>
</tr>
<tr>
<td>efficiency = toward-goal / total displacement</td>
<td>154</td>
<td>0.72</td>
<td>0.95</td>
<td>0.00</td>
<td>1.00</td>
</tr>
<tr>
<td>block rotation per bout (deg)</td>
<td>154</td>
<td>-0.0</td>
<td>0.0</td>
<td>-62.8</td>
<td>60.1</td>
</tr>
<tr>
<td></td>
<td>rotation</td>
<td>per bout (deg)</td>
<td>154</td>
<td>30.3</td>
<td>9.4</td>
</tr>
<tr>
<td>position error reduced per bout (px)</td>
<td>154</td>
<td>35.7</td>
<td>19.9</td>
<td>-1.0</td>
<td>102.1</td>
</tr>
<tr>
<td></td>
<td>angle error</td>
<td>reduced per bout (deg)</td>
<td>154</td>
<td>27.7</td>
<td>5.5</td>
</tr>
<tr>
<td>bout length (frames)</td>
<td>154</td>
<td>20</td>
<td>17</td>
<td>3</td>
<td>40</td>
</tr>
<tr>
<td>block speed on the release step (px/s, released bouts)</td>
<td>106</td>
<td>1.0</td>
<td>0.4</td>
<td>0.0</td>
<td>2.2</td>
</tr>
<tr>
<td>agent speed on the release step (px/s)</td>
<td>106</td>
<td>90.7</td>
<td>75.9</td>
<td>30.0</td>
<td>183.3</td>
</tr>
</tbody>
</table>
<p>What a bout achieves and when it is released (106 released bouts; the 48 bouts that run to the end of the episode are listed separately). <em>Push alignment</em> is the angle between the inward normal at the contact point (the push direction) and the CoG->goal direction.</p>
<table>
<thead>
<tr>
<th></th>
<th>n</th>
<th>mean</th>
<th>median</th>
<th>p10</th>
<th>p90</th>
</tr>
</thead>
<tbody>
<tr>
<td>push alignment at touch (deg, all bouts)</td>
<td>154</td>
<td>39.1</td>
<td>26.5</td>
<td>2.5</td>
<td>87.2</td>
</tr>
<tr>
<td>push alignment at release (deg)</td>
<td>106</td>
<td>92.1</td>
<td>88.8</td>
<td>47.6</td>
<td>152.0</td>
</tr>
<tr>
<td>push alignment at episode end (deg, final bouts)</td>
<td>48</td>
<td>22.2</td>
<td>13.9</td>
<td>2.8</td>
<td>53.2</td>
</tr>
<tr>
<td>position error at touch (px)</td>
<td>106</td>
<td>92.1</td>
<td>95.6</td>
<td>21.7</td>
<td>159.9</td>
</tr>
<tr>
<td>position error at release (px)</td>
<td>106</td>
<td>56.4</td>
<td>48.3</td>
<td>9.4</td>
<td>115.9</td>
</tr>
<tr>
<td></td>
<td>angle error</td>
<td>at touch (deg)</td>
<td>106</td>
<td>57.9</td>
<td>53.3</td>
</tr>
<tr>
<td></td>
<td>angle error</td>
<td>at release (deg)</td>
<td>106</td>
<td>18.7</td>
<td>5.3</td>
</tr>
<tr>
<td>position error at episode end (px, final bouts)</td>
<td>48</td>
<td>6.4</td>
<td>6.2</td>
<td>4.8</td>
<td>8.2</td>
</tr>
<tr>
<td></td>
<td>angle error</td>
<td>at episode end (deg, final bouts)</td>
<td>48</td>
<td>2.2</td>
<td>1.8</td>
</tr>
</tbody>
</table>
<p>Release happens with the push alignment above 45 deg in 91% of released bouts (above 60 deg in 83%), versus 34% of touches made with alignment above 45 deg. By release the bout has reduced the position error by more than 5 px in 71% of cases and the |angle error| by more than 5 deg in 70% (both 50%, neither 9%); the angle error changes sign during 24% of released bouts (overshoot).</p>
<p><img alt="push" src="../figures/human_pushing/push_dynamics.png" /></p>
<p><img alt="bouts" src="../figures/human_pushing/bout_release.png" /></p>
<h2 id="6-contact-selection-vs-error">6. Contact selection vs error</h2>
<p>Edge type chosen at deliberate touches, split by dominant error:</p>
<table>
<thead>
<tr>
<th>dominant error</th>
<th>touches</th>
<th>bar_top</th>
<th>bar_end</th>
<th>bar_bottom</th>
<th>stem_side</th>
<th>stem_tip</th>
</tr>
</thead>
<tbody>
<tr>
<td>position-dominant</td>
<td>120</td>
<td>28%</td>
<td>13%</td>
<td>7%</td>
<td>28%</td>
<td>24%</td>
</tr>
<tr>
<td>angle-dominant</td>
<td>29</td>
<td>21%</td>
<td>10%</td>
<td>17%</td>
<td>38%</td>
<td>14%</td>
</tr>
</tbody>
</table>
<p>Decision table fitted on all 149 deliberate touches of the 50 episodes. Events are mirrored across the stem axis so the goal is on the +x side (angle error and edge L/R flip with it); rows = direction to goal in the block frame, columns = angle error; cell = majority edge and its share:</p>
<p>| goal direction phi_g (deg) \ angle error (deg) | < -15 | |ang| <= 15 | > 15 |
|---|---|---|---|
| -90..-45 (bar-top side) | stem_L 4/7 | stem_tip 25/31 | stem_R 6/15 |
| -45..0 | stem_L 5/7 | stem_L 9/15 | bar_bottom_L 4/12 |
| 0..45 | bar_top 6/9 | stem_L 4/9 | stem_tip 1/2 |
| 45..90 (stem-tip side) | bar_top 5/8 | bar_top 20/25 | bar_top 5/9 |</p>
<p>Contact-point model: every candidate point q on the perimeter (every 5 px) gets a score <code>w1 * (-n_out . g) + w2 * (l/60) * angle_err + w3 * (-|l|/60)</code> where g is the unit direction to the goal in the block frame, l the lever arm of an inward push at q and angle_err in radians; the chosen point is a softmax draw (conditional logit) and the prediction is the argmax. <code>goal-normal</code> is the unfitted rule w = (1, 0, ~0): push where the inward normal points at the goal, as close to the CoG line as possible.</p>
<p>Fitted weights on all deliberate touches: w = (1.79, 2.97, 1.92).</p>
<table>
<thead>
<tr>
<th>rule</th>
<th>edge accuracy % (5-fold CV by episode, 50 eps)</th>
<th>edge accuracy % (external hold-out, 156 eps, 540 touches)</th>
<th>edge-type accuracy % (CV)</th>
<th>median perimeter error px (CV)</th>
</tr>
</thead>
<tbody>
<tr>
<td>majority edge</td>
<td>26.8</td>
<td>23.3</td>
<td>-</td>
<td>-</td>
</tr>
<tr>
<td>goal-normal (unfitted)</td>
<td>57.1</td>
<td>48.7</td>
<td>58.3</td>
<td>32.5</td>
</tr>
<tr>
<td>decision table (fitted)</td>
<td>54.5</td>
<td>56.3</td>
<td>-</td>
<td>-</td>
</tr>
<tr>
<td>contact-point logit (fitted)</td>
<td>56.4</td>
<td>50.7</td>
<td>56.4</td>
<td>24.5</td>
</tr>
</tbody>
</table>
<p>External hold-out for the logit: edge-type accuracy 51.1%, median perimeter error 45.0 px, p90 170.5 px. Chance level for 8 edges is 12.5%.</p>
<p><img alt="edge" src="../figures/human_pushing/edge_choice_vs_error.png" /></p>
<h2 id="7-design-implications-for-a-scripted-controller">7. Design implications for a scripted controller</h2>
<p>Each statement is read off the tables above (medians unless stated; p10-p90 in brackets). Speeds are agent-centre speeds at 10 Hz.</p>
<ol>
<li><strong>Approach speed and standoff.</strong> Do not back off and come in radially: 72% of no-contact motion is tangential (heading 45-135 deg to the outline). Orbit the block to the chosen edge at 123 px/s [57-210] with |acceleration| 305 px/s^2 [106-678], keeping a gap of 37 px [10-93] to the outline (max standoff between bouts 39 px [17-70]). The final radial APPROACH run is 4 frames [1-8] long, starts from a gap of 23 px [3-54], runs at 69 px/s [32-151] and arrives at 67 px/s [29-106]. The whole release-to-touch interval is 16 frames [7-25] and covers 98 px [25-182] of the 480 px perimeter.</li>
<li><strong>Contact normal alignment.</strong> Arrive along the inward normal: incidence 17 deg [2-51], 72% of touches within 30 deg, 7% glancing. Arrive behind the block as seen from the goal: bearing off the goal->block axis 24 deg [3-74] at touch, push alignment (push normal vs CoG->goal) 26 deg [3-87]. The edge is chosen by the goal geometry (item 7), not by proximity: the edge nearest the pusher at the previous release is touched only 9% of the time.</li>
<li><strong>Push depth and speed.</strong> Drive into the block at 48 px/s [20-89] (|acceleration| 88 px/s^2 [29-230], |jerk| 798 px/s^3 [252-2625]); the block follows at 32 px/s. One bout lasts 17 frames [3-40], displaces the CoG 44 px [4-125] with 0.95 of it along the goal direction, and rotates the block 9 deg [0-94]; it takes the position error from 96 to 48 px and the |angle error| from 53 to 5 deg (medians, released bouts). Keep the pushing bearing within 40 deg [4-107] of the goal->block axis during the push.</li>
<li><strong>Through the CoG or off-centre.</strong> Rotation follows the lever arm (slope 0.56 deg/s per px, r = 0.87, sign right in 100% of rotating PUSH frames). With |angle error| below 5 deg the human pushes through the CoG (median |lever| 5.7 px); at 10-20 deg the lever is 33.9 px and at 30-45 deg 42.5 px, with the corrective sign in 92.0% and 99.0% of frames. Split by dominant error: position-dominant pushes have median |lever| 25.2 px (40.4% within 15 px of the CoG line), angle-dominant pushes 42.5 px (11.5% through the CoG, 99.4% corrective). Rule: offset the contact point about 42.5 px from the CoG line on the side whose torque reduces the angle error once |angle error| exceeds ~10 deg; below 5 deg push through the CoG.</li>
<li><strong>Slide vs retreat.</strong> Slides are short in-contact re-aims: 2 frames [1-5] at 50 px/s [18-106], heading 64 deg off the inward normal, present in 60% of bouts; a PUSH run is followed by a SLIDE 71% of the time and a SLIDE returns to PUSH 51% of the time, so slide to re-aim while the push direction still points near the goal. Break contact when the push direction has rotated away from the goal: push alignment at release is 89 deg [48-152] (above 45 deg in 91% of releases) versus 26 deg at touch, typically after the |angle error| has been driven to 5 deg. Stop the block before leaving: block speed on the release step is 0.4 px/s (p90 2.2); the release step itself is taken at 76 px/s heading 124 deg off the inward normal (tangential in 61% of releases, straight out in 37%).</li>
<li><strong>Retreat distance.</strong> Retreat steps run at 106 px/s [32-250] but are few (170 frames, 2.7% of all): the pusher opens a gap of only 39 px [17-70] and then swings 89 deg [25-132] around the block (bearing change about the CoG, last contact -> next touch) before touching again 16 frames later.</li>
<li><strong>Choosing the next contact edge.</strong> Push where the inward normal points at the goal: the unfitted goal-normal rule picks the human's edge in 57% of held-out touches (CV) and 49% on the external hold-out (540 touches), against 27% / 23% for the majority edge and 12.5% chance. The fitted contact-point rule (w = 1.79, 2.97, 1.92: alignment, torque x angle error, off-centre penalty) reaches 56% / 51% and the mirrored decision table 54% / 56%; the torque term does not change which edge is chosen but moves the point along it (median perimeter error 24 px vs 32 px for goal-normal, CV). Read off the decision table (goal mirrored to the +x side, majority edge and its count): goal at -90..-45 (bar-top side) deg -> stem_L (4/7) for angle error < -15 deg, stem_tip (25/31) for |angle error| <= 15 deg, stem_R (6/15) for angle error > 15 deg; goal at -45..0 deg -> stem_L (5/7) for angle error < -15 deg, stem_L (9/15) for |angle error| <= 15 deg, bar_bottom_L (4/12) for angle error > 15 deg; goal at 0..45 deg -> bar_top (6/9) for angle error < -15 deg, stem_L (4/9) for |angle error| <= 15 deg, stem_tip (1/2) for angle error > 15 deg; goal at 45..90 (stem-tip side) deg -> bar_top (5/8) for angle error < -15 deg, bar_top (20/25) for |angle error| <= 15 deg, bar_top (5/9) for angle error > 15 deg. Position-dominant errors are served mostly by stem_side (28%) and bar_top (28%), angle-dominant errors by stem_side (38%) and bar_top (21%).</li>
</ol>
<p>Figures: <code>figures/human_pushing/</code>. Per-touch table: <code>figures/human_pushing/touch_events.csv</code>.</p>
<h2 id="8-which-findings-are-physics-and-which-are-the-human">8. Which findings are physics and which are the human</h2>
<p>The environment is a deterministic function of the action stream, so the split can be measured. Re-simulating the PD law (k_p=100, k_v=20, dt=0.01, 10 substeps) in free space from each recorded start, driven only by the recorded actions, reproduces the recorded agent path to 0.00 px (median and p90) on 3065 no-contact frames <strong>and on 3120 in-contact frames</strong>: the pusher is kinematic, the block never acts on it, and the whole agent trajectory is PD(mouse). The mouse stream is the only human quantity; everything else is a filter or a rigid-body response.</p>
<p><strong>Physics, fixed by the reference (would change on any other engine):</strong></p>
<ul>
<li>PD step response to a 100 px command: 29.6 / 61.6 / 80.8 / 95.6 px after 1-4 control steps, no overshoot, peak 320 px/s. A one-step lag and a ~10 px lead (median |target - agent| 10.1 px) are intrinsic, not chosen.</li>
<li>Smoothness: the PD halves the human's acceleration (mouse |accel| medians 224 push / 300 approach / 500 orbit / 1153 retreat px/s^2 vs agent 88 / 147 / 305 / 567) and quarters the jerk. About half of the "gentle" look is filter.</li>
<li>Push depth carries no force: the block is displaced by a moving wall. "Push at 48 px/s, block follows at 32" is kinematic slip, not contact mechanics.</li>
<li>Rotation vs lever arm (r = 0.87, 0.56 deg/s per px, sign correct in 100%) is rigid-body kinematics under the reference's understated moment (3000 vs 7875 geometric); the slope is a property of that typo and of damping = 0.</li>
<li>The block stopping on the release step (0.4 px/s) is damping = 0 annihilating velocity every substep.</li>
</ul>
<p><strong>Human, portable across engines (the decisions):</strong></p>
<table>
<thead>
<tr>
<th>decision</th>
<th>evidence</th>
</tr>
</thead>
<tbody>
<tr>
<td>speed by phase: slow in contact, fast when free</td>
<td>mouse speed medians 45 (push) / 41 (touch) / 51 (slide) vs 114 (orbit) / 184 (retreat) px/s; agent follows the mouse exactly (cos = 1.00 in every role)</td>
</tr>
<tr>
<td>orbit rather than retreat</td>
<td>72% of free motion tangential, retreat 2.7% of frames, standoff 37 px</td>
</tr>
<tr>
<td>head-on incidence</td>
<td>17 deg to the inward normal, 72% within 30 deg</td>
</tr>
<tr>
<td>edge selection by goal geometry</td>
<td>goal-normal rule 57% CV / 49% hold-out vs 9% nearest-edge</td>
</tr>
<tr>
<td>lever offset chosen by angle error</td>
<td>5.7 px at < 5 deg, 42.5 px at 30-45 deg, corrective sign 99%</td>
</tr>
<tr>
<td>release on geometry, not distance</td>
<td>push alignment 26 deg at touch, 89 deg at release</td>
</tr>
<tr>
<td>bout pacing</td>
<td>3 bouts of ~17 frames, 16 frames between</td>
</tr>
</tbody>
</table>
<p><strong>Consequences for the controller.</strong> Section 7's speeds are agent speeds; on a kinematic pusher the controller should emit <em>target-stream</em> speeds (the mouse numbers above) or it double-filters, and on a dynamic pusher the mapping differs again. The rotation gain is reference-only: close the loop on measured angle error instead of feeding forward 0.56 deg/s per px. The edge, lever, orbit, incidence and release rules are the transferable content of this analysis.</p></body></html> |